DOCSLIB.ORG

Lecture Notes for Optimization and Dynamics

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Lecture Notes for Optimization and Dynamics Lecture Notes for Optimization and Dynamics Joshua Maglione September 13, 2020 ii Copyright 2020, Joshua Maglione Based on notes of Terauds (2015) & Nilsson (2012, 2013). Contents 1 Introduction to dynamical systems1 1.1 Discrete and continuous time dynamical systems..................1 1.1.1 Newton{Raphson method...........................2 1.1.2 Exponential growth and decay........................3 1.1.3 The logistic map................................5 2 Discrete dynamical systems 13 2.1 Discrete autonomous dynamical systems....................... 14 2.2 Fixed points....................................... 18 2.3 Periodic orbits..................................... 22 3 One dimensional systems 25 3.1 Periodic points and Sharkovsky's Theorem...................... 30 4 Bifurcations in one dimensional systems 33 4.1 Saddle node bifurcation................................ 35 4.2 Transcritical bifurcation................................ 37 4.3 Pitchfork bifurcation.................................. 40 4.4 Period doubling bifurcation.............................. 44 4.5 Miscellaneous types of bifurcations.......................... 46 5 Linear discrete dynamical systems 49 5.1 Matrix norms...................................... 50 5.2 Jordan blocks...................................... 51 5.3 Stability criteria.................................... 53 6 Non-linear discrete dynamical systems 57 6.1 An introduction to Lyapunov exponents....................... 59 6.2 Some remarks on chaotic behavior.......................... 61 7 Ordinary differential equations and flows 65 7.1 Existence and uniqueness of solutions........................ 67 7.2 Fixed points and autonomous systems........................ 69 7.3 Flows.......................................... 70 8 Linear ODEs of higher dimension 73 8.1 Matrix exponentials.................................. 73 8.2 Characteristic polynomials............................... 77 8.3 Solving linear ODEs.................................. 80 9 Non-linear ODEs 89 9.1 Lyapunov functions................................... 91 iii iv CONTENTS 10 Dynamic Programming 97 10.1 The Principle of Optimality.............................. 99 Bibliography 101 Index 103 Chapter 1 Introduction to dynamical systems The fundamental idea of the course is to use mathematics to make predictions about the future. We do this all the time in many different contexts. For example: (Finance) Trajectory of interest rates, inflation rate, and GDP, (Finance) Call options for stocks, (Biology) Spread of infectious diseases, (Biology) Population growth, (Physics) Heat transfer, (Physics) Positions and velocities of planets and stars. These are examples of different kinds of systems. Definition 1.1. A system is a set of measurable quantities, and a dynamical system is a system that changes over time. In order to analyze systems, we build (mathematical) models. There is no such thing as a perfect model! But we can build some incredibly accurate models for some systems. Definition 1.2. The state of a system is the set of values describing the system at that time. If we know the state of a system at some time t = t0, the model allows us to predict the state of the system at some future time t = tn. Every dynamical system consists of two parts: 1. the state space: the set of all possible states of the system and 2. the time evolution rule: the rule (function) that describes how the states of the system change over time. A time evolution rule may take various forms, but we often try to convert it to one that gives the state of the system at a general time t in terms of an initial state. 1.1 Discrete and continuous time dynamical systems Dynamical systems are classified as either discrete or continuous based on the nature of how the system changes over time. Definition 1.3. A discrete (time) dynamical system is a dynamical system that changes state in discrete time steps. For example at t0; t1; t2;::: . 1 2 CHAPTER 1. INTRODUCTION TO DYNAMICAL SYSTEMS Consider, for example, the balance of a savings account where interest is compounded monthly. Discrete systems can be described by difference equations or recurrence relations. Definition 1.4. A continuous (time) dynamical system is a dynamical system that changes state continuously over time. Consider, for example, the position of a swinging pendulum, or the value of a commodity. Such systems can be described by differential equations. We now consider some more detailed examples of dynamical systems. 1.1.1 Newton{Raphson method The Newton{Raphson method is an iterative numerical method for finding real roots of differ- entiable functions. It was first developed by Newton in Method of Fluxions around 1671 but published after Raphson's version around 1690, which is simpler than Newton's. Let f : R ! R be a differentiable function with a real rootx ¯. Given a point xn 2 R, we denote by Tn the tangent line to the curve f at the point (xn; f(xn)). We define a new point xn+1 2 R to be the intersection of this tangent line with the x-axis. Recall, the slope of the tangent line is Tn(x) is equal to the slope of the curve f at (xn; f(xn)). With Tn(x) = mx + c, we have 0 m = f (xn); f(xn) = mxn + c; 0 which gives c = f(xn) − f (xn)xn. The equation of the tangent line is therefore 0 0 Tn(x) = f (xn)x + f(xn) − f (xn)xn 0 = f (xn)(x − xn) + f(xn): This line intersects the x-axis when Tn(x) = 0. This occurs when f(xn) x = xn − 0 : f (xn) Thus, we set f(xn) xn+1 = xn − 0 ; (1.1) f (xn) and we iterate this process to numerically approximate the root of f \near" some initial guess. y x2 x1 x0 x Figure 1.1: Illustrating the first few iterations of the Newton{Raphson method. So how is this a dynamical system? The quantity that we are describing is the approximate value of the root. The state of the system at time (step) n is xn, which may be any real number. 1.1. DISCRETE AND CONTINUOUS TIME DYNAMICAL SYSTEMS 3 So the state space of the system is R. The dynamical system is discrete since there is one step after another, and the time evolution rule is given by equation (1.1). We start the iteration with an initial value x0, the initial state of the system. This value determines whether or not the state converges to the actual root of the function. There are continuous functions for which no starting value will converge; see the exercises. 1.1.2 Exponential growth and decay Sometimes we may choose whether to model a dynamical system with discrete or continuous time. In the next two examples, we compare discrete and continuous models of the same system concerning exponential growth/decay. This system is one of the most basic and fundamental systems. + First, we consider the discrete system. Let a 2 R , x0 2 R, and for all n 2 N, xn+1 = axn; (1.2) By iterating, we get x1 = ax0 2 x2 = ax1 = a x0 3 x3 = ax2 = a x0 . n xn = axn−1 = a x0: Now we explore the qualitative behavior of this system for different values of the constant a. Case 1: a = 1. Then for all n 2 N, xn = x0, so the system is constant. This is uninteresting. Case 2: a > 1. We consider three different cases based on the initial value x0. If x0 = 0, then xn = 0 for all n 2 N. We say that x = 0 is a fixed point. If x0 > 0, then xn > xn−1 for all n. That is, the sequence (xn)n2N is monotone increasing and xn ! 1 as n ! 1. If x0 < 0, then xn < xn−1 for all n. That is, the sequence (xn)n2N is monotone decreasing and xn ! −∞ as n ! 1. We will discuss this in more detail later, but for now we call x = 0 an unstable fixed point since the system is \going away" from x = 0. If the system has the state x = 0, then it will stay there. If we start the system just a small distance away from x = 0, then the value moves further away from this point. Case 3: 0 < a < 1. As in the case above, we consider three different situations. If x0 = 0, then xn = 0 for all n 2 N. If x0 > 0, then 0 < xn < xn−1 for all n, and xn ! 0 as n ! 1. If x0 < 0, then 0 > xn > xn−1 for all n, and xn ! 0 as n ! 1. In this case, we say that x = 0 is a stable fixed point since the system is \coming towards" x = 0. We can perturb the starting point away from the state x = 0, and the system will evolve back to the state x = 0. 4 CHAPTER 1. INTRODUCTION TO DYNAMICAL SYSTEMS Remark 1.5. The equation xn+1 = axn is called a difference equation as we can write it in the form of a difference: xn+1 − xn = axn − xn = (a − 1)xn = λxn: (1.3) | {z } difference The change in state|that is, the difference between the next state xn+1 and the current state xn|depends on the constant λ = a − 1 and the current state xn. Now we turn to the continuous version of the exponential growth/decay system. We will denote the state of the system at time t by x(t) and the initial state, as above, by x0 = x(0). The change in the state of the system still depends on the current state and a constant, but as the change is considered to be continuous, we use a derivative rather than a difference. The growth/decay equation (1.3) then becomes dx x0(t) := = λ x(t) : (1.4) dt To convert the time evolution rule into a more \useful" form, we solve the differential equation. Equation (1.4) becomes x0(t) = λ. (1.5) x(t) Now we can integrate both sides of (1.5) with respect to t.
Load more

Recommended publications
  • Linear Systems
    Linear Systems Professor Yi Ma Professor Claire Tomlin GSI: Somil Bansal Scribe: Chih-Yuan Chiu Department of Electrical Engineering and Computer Sciences University of California, Berkeley Berkeley, CA, U.S.A. May 22, 2019 2 Contents Preface 5 Notation 7 1 Introduction 9 1.1 Lecture 1 . .9 2 Linear Algebra Review 15 2.1 Lecture 2 . 15 2.2 Lecture 3 . 22 2.3 Lecture 3 Discussion . 28 2.4 Lecture 4 . 30 2.5 Lecture 4 Discussion . 35 2.6 Lecture 5 . 37 2.7 Lecture 6 . 42 3 Dynamical Systems 45 3.1 Lecture 7 . 45 3.2 Lecture 7 Discussion . 51 3.3 Lecture 8 . 55 3.4 Lecture 8 Discussion . 60 3.5 Lecture 9 . 61 3.6 Lecture 9 Discussion . 67 3.7 Lecture 10 . 72 3.8 Lecture 10 Discussion . 86 4 System Stability 91 4.1 Lecture 12 . 91 4.2 Lecture 12 Discussion . 101 4.3 Lecture 13 . 106 4.4 Lecture 13 Discussion . 117 4.5 Lecture 14 . 120 4.6 Lecture 14 Discussion . 126 4.7 Lecture 15 . 127 3 4 CONTENTS 4.8 Lecture 15 Discussion . 148 5 Controllability and Observability 157 5.1 Lecture 16 . 157 5.2 Lecture 17 . 163 5.3 Lectures 16, 17 Discussion . 176 5.4 Lecture 18 . 182 5.5 Lecture 19 . 185 5.6 Lecture 20 . 194 5.7 Lectures 18, 19, 20 Discussion . 211 5.8 Lecture 21 . 216 5.9 Lecture 22 . 222 6 Additional Topics 233 6.1 Lecture 11 . 233 6.2 Hamilton-Jacobi-Bellman Equation . 254 A Appendix to Lecture 12 259 A.1 Cayley-Hamilton Theorem: Alternative Proof 1 .
    [Show full text]
  • 3.3 Diagonalization and Eigenvalues
    3.3. Diagonalization and Eigenvalues 171 n 1 3 0 1 Exercise 3.2.33 Show that adj (uA)= u − adj A for all 1 Exercise 3.2.28 If A− = 0 2 3 find adj A. n n A matrices . 3 1 1 × − Exercise 3.2.34 Let A and B denote invertible n n ma- Exercise 3.2.29 If A is 3 3 and det A = 2, find × 1 × trices. Show that: det (A− + 4 adj A). 0 A Exercise 3.2.30 Show that det = det A det B a. adj (adj A) = (det A)n 2A (here n 2) [Hint: See B X − ≥ Example 3.2.8.] when A and B are 2 2. What ifA and Bare 3 3? × × 0 I [Hint: Block multiply by .] b. adj (A 1) = (adj A) 1 I 0 − − Exercise 3.2.31 Let A be n n, n 2, and assume one c. adj (AT ) = (adj A)T column of A consists of zeros.× Find≥ the possible values of rank (adj A). d. adj (AB) = (adj B)(adj A) [Hint: Show that AB adj (AB)= AB adj B adj A.] Exercise 3.2.32 If A is 3 3 and invertible, compute × det ( A2(adj A) 1). − − 3.3 Diagonalization and Eigenvalues The world is filled with examples of systems that evolve in time—the weather in a region, the economy of a nation, the diversity of an ecosystem, etc. Describing such systems is difficult in general and various methods have been developed in special cases. In this section we describe one such method, called diag- onalization, which is one of the most important techniques in linear algebra.
    [Show full text]
  • Linear Dynamics: Clustering Without Identification
    Linear Dynamics: Clustering without identification Chloe Ching-Yun Hsuy Michaela Hardty Moritz Hardty University of California, Berkeley Amazon University of California, Berkeley Abstract for LDS parameter estimation, but it is inherently non-convex and can often get stuck in local min- Linear dynamical systems are a fundamental ima [Hazan et al., 2018]. Even when full system iden- and powerful parametric model class. How- tification is hard, is there still hope to learn meaningful ever, identifying the parameters of a linear information about linear dynamics without learning all dynamical system is a venerable task, per- system parameters? We provide a positive answer to mitting provably efficient solutions only in this question. special cases. This work shows that the eigen- We show that the eigenspectrum of the state-transition spectrum of unknown linear dynamics can be matrix of unknown linear dynamics can be identified identified without full system identification. without full system identification. The eigenvalues of We analyze a computationally efficient and the state-transition matrix play a significant role in provably convergent algorithm to estimate the determining the properties of a linear system. For eigenvalues of the state-transition matrix in example, in two dimensions, the eigenvalues determine a linear dynamical system. the stability of a linear dynamical system. Based on When applied to time series clustering, the trace and the determinant of the state-transition our algorithm can efficiently cluster multi- matrix, we can classify a linear system as a stable dimensional time series with temporal offsets node, a stable spiral, a saddle, an unstable node, or an and varying lengths, under the assumption unstable spiral.
    [Show full text]
  • Binet-Cauchy Kernels on Dynamical Systems and Its Application to the Analysis of Dynamic Scenes ∗
    Binet-Cauchy Kernels on Dynamical Systems and its Application to the Analysis of Dynamic Scenes ∗ S.V.N. Vishwanathan Statistical Machine Learning Program National ICT Australia Canberra, 0200 ACT, Australia Tel: +61 (2) 6125 8657 E-mail: [email protected] Alexander J. Smola Statistical Machine Learning Program National ICT Australia Canberra, 0200 ACT, Australia Tel: +61 (2) 6125 8652 E-mail: [email protected] Ren´eVidal Center for Imaging Science, Department of Biomedical Engineering, Johns Hopkins University 308B Clark Hall, 3400 N. Charles St., Baltimore MD 21218 Tel: +1 (410) 516 7306 E-mail: [email protected] October 5, 2010 Abstract. We propose a family of kernels based on the Binet-Cauchy theorem, and its extension to Fredholm operators. Our derivation provides a unifying framework for all kernels on dynamical systems currently used in machine learning, including kernels derived from the behavioral framework, diffusion processes, marginalized kernels, kernels on graphs, and the kernels on sets arising from the subspace angle approach. In the case of linear time-invariant systems, we derive explicit formulae for computing the proposed Binet-Cauchy kernels by solving Sylvester equations, and relate the proposed kernels to existing kernels based on cepstrum coefficients and subspace angles. We show efficient methods for computing our kernels which make them viable for the practitioner. Besides their theoretical appeal, these kernels can be used efficiently in the comparison of video sequences of dynamic scenes that can be modeled as the output of a linear time-invariant dynamical system. One advantage of our kernels is that they take the initial conditions of the dynamical systems into account.
    [Show full text]
  • Arxiv:1908.01039V3 [Cs.LG] 29 Feb 2020 LDS Eigenvalues
    Linear Dynamics: Clustering without identification Chloe Ching-Yun Hsuy Michaela Hardty Moritz Hardty University of California, Berkeley Amazon University of California, Berkeley Abstract for LDS parameter estimation, but it is inherently non-convex and can often get stuck in local min- Linear dynamical systems are a fundamental ima [Hazan et al., 2018]. Even when full system iden- and powerful parametric model class. How- tification is hard, is there still hope to learn meaningful ever, identifying the parameters of a linear information about linear dynamics without learning all dynamical system is a venerable task, per- system parameters? We provide a positive answer to mitting provably efficient solutions only in this question. special cases. This work shows that the eigen- We show that the eigenspectrum of the state-transition spectrum of unknown linear dynamics can be matrix of unknown linear dynamics can be identified identified without full system identification. without full system identification. The eigenvalues of We analyze a computationally efficient and the state-transition matrix play a significant role in provably convergent algorithm to estimate the determining the properties of a linear system. For eigenvalues of the state-transition matrix in example, in two dimensions, the eigenvalues determine a linear dynamical system. the stability of a linear dynamical system. Based on When applied to time series clustering, the trace and the determinant of the state-transition our algorithm can efficiently cluster multi- matrix, we can classify a linear system as a stable dimensional time series with temporal offsets node, a stable spiral, a saddle, an unstable node, or an and varying lengths, under the assumption unstable spiral.
    [Show full text]
  • Linear Dynamical Systems As a Core Computational Primitive
    Linear Dynamical Systems as a Core Computational Primitive Shiva Kaul Computer Science Department Carnegie Mellon University Pittsburgh, PA 15213 [email protected] Abstract Running nonlinear RNNs for T steps takes Ω(T ) time. Our construction, called LDStack, approximately runs them in O(log T ) parallel time, and obtains arbitrarily low error via repetition. First, we show nonlinear RNNs can be approximated by a stack of multiple-input, multiple-output (MIMO) LDS. This replaces nonlinearity across time with nonlinearity along depth. Next, we show that MIMO LDS can be approximated by an average or a concatenation of single-input, multiple-output (SIMO) LDS. Finally, we present an algorithm for running (and differentiating) SIMO LDS in O(log T ) parallel time. On long sequences, LDStack is much faster than traditional RNNs, yet it achieves similar accuracy in our experiments. Furthermore, LDStack is amenable to linear systems theory. Therefore, it improves not only speed, but also interpretability and mathematical tractability. 1 Introduction Nonlinear RNNs have two crucial shortcomings. The first is computational: running an RNN for T steps is a sequential operation which takes Ω(T ) time. The second is analytical: it is challenging to gain intuition about the behavior of a nonlinear RNN, and even harder to prove this behavior is desirable. These shortcomings have motivated practitioners to abandon RNNs altogether and to model time series by other means. These include hierarchies of (dilated) convolutions [Oord et al., 2016, Gehring et al., 2017] and attention mechanisms which are differentiable analogues of key-value lookups [Bahdanau et al., 2014, Vaswani et al., 2017].
    [Show full text]
  • Dynamical Systems and Matrix Algebra
    Dynamical Systems and Matrix Algebra K. Behrend August 12, 2018 Abstract This is a review of how matrix algebra applies to linear dynamical systems. We treat the discrete and the continuous case. 1 Contents Introduction 4 1 Discrete Dynamical Systems 4 1.1 A Markov Process . 4 A migration example . 4 Translating the problem into matrix algebra . 4 Finding the equilibrium . 7 The line of fixed points . 9 1.2 Fibonacci's Example . 11 Description of the dynamical system . 11 Model using rabbit vectors . 12 Starting the analysis . 13 The method of eigenvalues . 13 Finding the eigenvalues . 17 Finding the eigenvectors . 18 Qualitative description of the long-term behaviour . 20 The second eigenvalue . 20 Exact solution . 22 Finding the constants . 24 More detailed analysis . 25 The Golden Ratio . 26 1.3 Predator-Prey System . 27 Frogs and flies . 27 The model . 28 Solving the system . 29 Discussion of the solution . 31 The phase portrait . 32 The method of diagonalization . 34 Concluding remarks . 39 1.4 Summary of the Method . 40 The method of undetermined coefficients . 41 The method of diagonalization . 41 The long term behaviour . 42 1.5 Worked Examples . 44 A 3-dimensional dynamical system . 44 The powers of a matrix . 49 2 1.6 More on Markov processes . 53 1.7 Exercises . 59 2 Continuous Dynamical Systems 60 2.1 Flow Example . 60 2.2 Discrete Model . 61 2.3 Refining the Discrete Model . 62 2.4 The continuous model . 63 2.5 Solving the system of differential equations . 64 One homogeneous linear differential equation . 64 Our system of linear differential equations .
    [Show full text]
  • Lecture Notes for EE263
    Lecture Notes for EE263 Stephen Boyd Introduction to Linear Dynamical Systems Autumn 2007-08 Copyright Stephen Boyd. Limited copying or use for educational purposes is fine, but please acknowledge source, e.g., “taken from Lecture Notes for EE263, Stephen Boyd, Stanford 2007.” Contents Lecture 1 – Overview Lecture 2 – Linear functions and examples Lecture 3 – Linear algebra review Lecture 4 – Orthonormal sets of vectors and QR factorization Lecture 5 – Least-squares Lecture 6 – Least-squares applications Lecture 7 – Regularized least-squares and Gauss-Newton method Lecture 8 – Least-norm solutions of underdetermined equations Lecture 9 – Autonomous linear dynamical systems Lecture 10 – Solution via Laplace transform and matrix exponential Lecture 11 – Eigenvectors and diagonalization Lecture 12 – Jordan canonical form Lecture 13 – Linear dynamical systems with inputs and outputs Lecture 14 – Example: Aircraft dynamics Lecture 15 – Symmetric matrices, quadratic forms, matrix norm, and SVD Lecture 16 – SVD applications Lecture 17 – Example: Quantum mechanics Lecture 18 – Controllability and state transfer Lecture 19 – Observability and state estimation Lecture 20 – Some final comments Basic notation Matrix primer Crimes against matrices Solving general linear equations using Matlab Least-squares and least-norm solutions using Matlab Exercises EE263 Autumn 2007-08 Stephen Boyd Lecture 1 Overview • course mechanics • outline & topics • what is a linear dynamical system? • why study linear systems? • some examples 1–1 Course mechanics • all class
    [Show full text]
  • Dynamical Systems Dennis Pixton
    Dynamical Systems Version 0.2 Dennis Pixton E-mail address: [email protected] Department of Mathematical Sciences Binghamton University Copyright 2009{2010 by the author. All rights reserved. The most current version of this book is available at the website http://www.math.binghamton.edu/dennis/dynsys.html. This book may be freely reproduced and distributed, provided that it is reproduced in its entirety from one of the versions which is posted on the website above at the time of reproduction. This book may not be altered in any way, except for changes in format required for printing or other distribution, without the permission of the author. Contents Chapter 1. Discrete population models 1 1.1. The basic model 1 1.2. Discrete dynamical systems 3 1.3. Some variations 4 1.4. Steady states and limit states 6 1.5. Bounce graphs 8 Exercises 12 Chapter 2. Continuous population models 15 2.1. The basic model 15 2.2. Continuous dynamical systems 18 2.3. Some variations 19 2.4. Steady states and limit states 26 2.5. Existence and uniqueness 29 Exercises 34 Chapter 3. Discrete Linear models 37 3.1. A stratified population model 37 3.2. Matrix powers, eigenvalues and eigenvectors. 40 3.3. Non negative matrices 44 3.4. Networks; more examples 45 3.5. Google PageRank 51 3.6. Complex eigenvalues 55 Exercises 60 Chapter 4. Linear models: continuous version 65 4.1. The exponential function 65 4.2. Some models 72 4.3. Phase portraits 78 Exercises 87 Chapter 5. Non-linear systems 90 5.1.
    [Show full text]
  • An Elementary Introduction to Linear Dynamical System
    An elementary introduction to linear dynamical system Bijan Bagchi Department of Applied Mathematics University of Calcutta E-mail : bbagchi123@rediffmail.com 18.01.2012 1 What is a dynamical system (DS)? Mathematically, a DS deals with an initial value problem of the type d−!x −! = f (t; −!x ; u) dt −! k where x denotes a vector having components (x1; x2; :::; xk) 2 R ; t is the −! −! time, f is a vector flow : f = (f1; f2; :::; fk); fug is a set of auxiliary objects −! and there exist a set of initial conditions x0 = [x1(0); x2(0); :::; xk(0)]. Thus we write 0 1 0 1 0 1 x1 f1 x1(0) B C B C B C B C B C B C B x2 C B f2 C B x2(0) C B C B C B C B C B C B C −! B : C −! B : C −! B : C x = B C, f = B C, x0 = B C. B C B C B C B : C B : C B : C B C B C B C B C B C B C B : C B : C B : C @ A @ A @ A xk fk xk(0) The case k = 1 is trivial implying a scalar equation with the solution Z t x = x0 + f(s; x; u) ds: 0 The case k = 2 has the form (with no explicit presence of t) x_i = fi(x1; x2) i = 1; 2 where we also assume fi(x1; x2) to be continuously differentiable in the neigh- borhood of [x1(0); x2(0)].
    [Show full text]
  • Lecture Notes for Introduction to Dynamical Systems: CM131A 2018-2019
    Lecture notes for Introduction to Dynamical Systems: CM131A Based on notes by A. Annibale, R. K¨uhnand H.C. Rae 2018-2019 Lecturer: G. Watts 1 Contents 1 Overview of the course 4 1.1 Revision Exercises . .9 I Differential Equations 10 2 First order Differential Equations 11 2.1 Basic Ideas . 11 2.2 First order differential equations . 12 2.3 General solution of specific equations . 13 2.3.1 First order, explicit . 13 2.3.2 First order, variables separable . 14 2.3.3 First order, linear . 16 2.3.4 First order, homogeneous . 18 2.4 Initial value problems . 20 2.5 Existence and Uniqueness of Solutions | Picard's theorem . 23 2.5.1 Picard iterates . 24 2.6 Exercises . 28 3 Second order Differential Equations 31 3.1 Second order differential equations . 31 3.1.1 Second order linear, with constant coefficients . 32 3.2 Existence and Uniqueness of Solutions | Picard's theorem . 38 3.3 Exercises . 41 II Dynamical Systems 44 4 Introduction to Dynamical Systems 45 Version of Mar 14, 2019 2 5 First Order Autonomous Systems 50 5.1 Trajectories, orbits and phase portraits . 50 5.2 Termination of Motion . 55 5.3 Estimating times of Motion . 60 5.4 Stability | A More General Discussion . 66 5.4.1 Stability of Fixed Points . 66 5.4.2 Structural Stability . 68 5.4.3 Stability of Motion . 71 5.5 Asymptotic Analysis . 72 5.5.1 Asymptotic Analysis and Dynamical Systems . 74 5.6 Exercises . 80 6 Second Order Autonomous Systems 85 6.1 Phase Space and Phase Portraits .
    [Show full text]
  • Mathematical Description of Linear Dynamical Systems* R
    J.S.I.A.M. CONTROI Ser. A, Vol. 1, No. Printed in U.,q.A., 1963 MATHEMATICAL DESCRIPTION OF LINEAR DYNAMICAL SYSTEMS* R. E. KALMAN Abstract. There are two different ways of describing dynamical systems: (i) by means of state w.riables and (if) by input/output relations. The first method may be regarded as an axiomatization of Newton's laws of mechanics and is taken to be the basic definition of a system. It is then shown (in the linear case) that the input/output relations determine only one prt of a system, that which is completely observable and completely con- trollable. Using the theory of controllability and observability, methods are given for calculating irreducible realizations of a given impulse-response matrix. In par- ticular, an explicit procedure is given to determine the minimal number of state varibles necessary to realize a given transfer-function matrix. Difficulties arising from the use of reducible realizations are discussed briefly. 1. Introduction and summary. Recent developments in optimM control system theory are bsed on vector differential equations as models of physical systems. In the older literature on control theory, however, the same systems are modeled by ransfer functions (i.e., by the Laplace trans- forms of the differential equations relating the inputs to the outputs). Two differet languages have arisen, both of which purport to talk about the same problem. In the new approach, we talk about state variables, tran- sition equations, etc., and make constant use of abstract linear algebra. In the old approach, the key words are frequency response, pole-zero pat- terns, etc., and the main mathematical tool is complex function theory.
    [Show full text]